Design of FLAT-SUGAR:Randomized Trial of PrandialInsulin Versus Prandial GLP-1Receptor Agonist Together WithBasal Insulin and Metformin forHigh-Risk Type 2 DiabetesDiabetes Care 2015;38:1558–1566 | DOI: 10.2337/dc14-2689

OBJECTIVE

Glycemic variability may contribute to adverse medical outcomes of type 2 di-abetes, but prior therapies have had limited success in controlling glycemic fluc-tuations, and the hypothesis has not been adequately tested.

RESEARCH DESIGN AND METHODS

People with insulin-requiring type 2 diabetes and high cardiovascular risk wereenrolled during a run-in period on basal-bolus insulin (BBI), and 102 were ran-domized to continued BBI or to basal insulin with a prandial GLP-1 receptoragonist (GLIPULIN) group, each seeking to maintain HbA1c levels between 6.7%and 7.3% (50–56 mmol/mol) for 6 months. The primary outcome measure wasglycemic variability assessed by continuous glucose monitoring; other measureswere HbA1c, weight, circulating markers of inflammation and cardiovascular risk,albuminuria, and electrocardiographic patterns assessed by Holter monitoring.

RESULTS

At randomization, the mean age of the population was 62 years, median durationof diabetes 15 years, mean BMI 34 kg/m2, and mean HbA1c 7.9% (63 mmol/mol).Thirty-three percent had a prior cardiovascular event, 18% hadmicroalbuminuria,and 3% had macroalbuminuria. At baseline, the continuous glucose monitoringcoefficient of variation for glucose levels was similar in both groups.

CONCLUSIONS

FLAT-SUGAR is a proof-of-concept study testing whether, in a population of indi-viduals with type 2 diabetes and high cardiovascular risk, the GLIPULIN regimencan limit glycemic variability more effectively than BBI, reduce levels of cardio-vascular risk markers, and favorably alter albuminuria and electrocardiographicpatterns. We successfully randomized a population that has sufficient power toanswer the primary question, address several secondary ones, and complete theprotocol as designed.

Corresponding author: Jeffrey L. Probstfield,jeffprob@uw.edu.

Received 14 November 2014 and accepted 20May 2015.

Clinical trial reg. no. NCT01524705, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc14-2689/-/DC1.

*The members of the Writing Committee of theFLAT-SUGAR Trial are listed in the APPENDIX.

© 2015 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered.

See accompanying articles, pp. 1610and 1615.

The FLAT-SUGAR Trial Investigators*

1558 Diabetes Care Volume 38, August 2015

CARDIOVASC

ULA

RANDMETABOLICRISK

Improving mean levels of glycemic con-trol, as judged by assays for glycatedhemoglobin (HbA1c), has been shownto decrease risks of microvascular com-plications and cardiovascular disease(CVD) events for people with type 1and type 2 diabetes (1,2). On the basisof these trials, HbA1c levels ,6.5% or7% (47 or 53 mmol/mol) are proposedas standards for glycemic control (3,4).However, three subsequent trials sug-gested that targeting HbA1c to this rangemay not lead to improved outcomes forpeople with long-standing type 2 dia-betes (5–7). Of particular concern isthe increase in CVD and all-cause mor-tality accompanying an intensive glyce-mic treatment strategy in the Action toControl Cardiovascular Risk in Diabetes(ACCORD) trial (6). The reasons for fail-ure of the HbA1c-based glycemic controlstrategies used in these more recentstudies are not clear, but one possibilityis that HbA1c levels do not assess vari-ability in plasma glucose levels; that is,HbA1c does not directly reflect the fre-quency or severity of either hypoglyce-mic events or hyperglycemic incrementsafter meals, both of which may causephysiologic changes that contribute tolong-term risks. Possible mechanisms forsuch effects include systemic inflamma-tion (8,9), oxidative stress (10), endothe-lial dysfunction (11,12), intimal-medialthickening (13), and cardiac ischemia orarrhythmias (14–16). Considerable litera-ture linking various CVD risk markers withglycemic variability has been published(17,18).Directly testing the hypothesis that gly-

cemic variability contributes to adversemedical outcomes has been difficult dueto limitations in available treatments anddiagnostic tools. Other thana-glucosidaseinhibitors, oral therapies have relativelyweak effects on postprandial hypergly-cemia. Of note, after failure of othertreatments in people with long-standingtype 2 diabetes, management of post-prandial hyperglycemia with preprandialinjections of rapid-acting insulin addedto basal insulin is challenging and oftenineffective (19,20). Efforts to maximizemealtimeglycemic control by thesemeanscommonly fail to provide adequate con-trol of postprandial glycemic incrementswhile causing frequent hypoglycemiaand weight gain.However, new treatments for type 2

diabetes have been introduced. The

shorter-acting GLP-1 receptor agonists(GLP-1RAs) exenatide and lixisenatideare of particular interest because oftheir ability to blunt postprandial glyce-mic increments (mainly by slowing gas-tric emptying) while also potentiatingendogenous insulin secretion and sup-pressing inappropriate elevations of glu-cagon (21–25). Studies have shown thatthese agents, when used in combinationwith basal insulin and metformin, canflatten the postprandial glucose profilewithout increasing the risks of hypogly-cemia and weight gain often associatedwith prandial insulin treatment (26–29).In addition, improved devices for con-tinuous glucosemonitoring (CGM) allowfor more complete assessment of 24-hglycemic patterns (30–33).

The FLuctuATion reduction with inSUlinand Glp-1 Added togetheR (FLAT-SUGAR)study is using these methods to examinethe hypothesis that glycemic variability isan important component of diabetes con-trol that may influence diabetes compli-cations. Three specific study questionsare addressed. On a background of met-formin and long-acting insulin therapyand with similar levels of HbA1c attained,does mealtime treatment with exena-tide compared with rapid-acting insulin1) reduce overall glycemic variability, 2)reduce circulating levels of CVDmarkers,and 3) reduce albuminuria or Holter-monitored electrocardiographic (ECG)abnormalities, two measures of tissuefunction relevant to diabetes? The re-sults of the present proof-of-conceptstudy are intended to guide the designof a larger, medical outcomes–driventrial testing a treatment strategy thatreduces glucose variability better thantraditional basal-bolus insulin (BBI) ther-apy. Because of this long-term objective,the population studied in FLAT-SUGARwas intended to be similar to that of theACCORD trial.

RESEARCH DESIGN AND METHODS

Study OrganizationFLAT-SUGAR is an Internet-based study(https://ccct.sph.uth.tmc.edu/flatsugar)with a public section as well as a password-protected section for use by clinical sitesfor randomization and data entry. Allstudy documents are in the password-protected area of the Web site.

Logistical components of the trial in-clude, besides the Clinical Trials ServiceUnit at the University of Washington in

Seattle, Washington, a Data CoordinatingCenter at the University of Texas Schoolof Public Health in Houston, Texas; a DrugDistribution Center at the Veterans Af-fairs Cooperative Studies Program in Al-buquerque, New Mexico; and a CoreLaboratory and a Holter ECG Readingand Analysis Center at the University ofWashington. The organizational compo-nents of the study and their interrelation-ships are described in SupplementaryFig. 1. Clinical site investigators with ex-pertise in intensive management of dia-betes, including the use of CGM, and inclinical studymanagement agreed to par-ticipate in this demanding protocol.These features permitted the ambitiousgoal of testing proof of concept for threeseparate study questions.

Study DesignThe study is amulticenter comparison oftwo glycemic treatment strategies andcomprised a screening period, an 8–12-week open-label run-in period, and a26-week randomized, open-label treat-ment period (Supplementary Fig. 2).

Study Population Selected atScreeningA full description of inclusion and exclu-sion criteria is shown in SupplementaryTable 1. Briefly, we identified an ACCORD-like population required to have type 2diabetes and to need insulin therapy.Candidates for enrollment had to haveeither clinical CVD or evidence of in-creased CVD risk and to be clinically sta-ble for the 3 months before enrollment.A subsamplewith some renal dysfunctionwas identified. Participants were pro-vided no monetary incentive except asmall reimbursement for transportation(gas mileage or fares), parking, and mealsat the time of visits (total $10–50).

Run-in PeriodThe run-in procedure provided a stableclinical baseline, familiarized candidatesto be randomized (CTR) with the studyprocedures, and identified those un-able to adhere to study requirements.All CTR received instruction on an Amer-ican Heart Association/American Diabe-tes Association meal plan and desirableexercises. If they had been using car-bohydrate counting before the study,CTR were allowed to continue to doso as long as they met the HbA1c goal.During this 8–12-week period, enrolledCTR used BBI therapy given by separate

care.diabetesjournals.org The FLAT-SUGAR Trial Investigators 1559

injections with metformin and per-formed self-monitoring of blood glucose(SMBG) using the Bayer Contour meterthree to four times daily as needed. AllCTR were placed on basal insulin treat-ment using once- or twice-daily insulinglargine. If not previously taking metfor-min, CTR were required to start it withtitration from 500 to 2,000 mg daily astolerated. Medication dosages were ad-justed by the investigators with a goal tomaintain HbA1c levels in the 6.7–8.0%(50–64 mmol/mol) range. All CTR wererequired to take one of three insulinrapid-acting analogs (RAAs): aspart,glulisine, or lispro. CGM (DexcomSEVEN PLUS or G4), which was maskedto CTR and investigators, and ambula-tory ECG monitoring (Medicomp Holter)were performed for 7–10 days after atleast 6 weeks of stable glycemic control.At the end of the run-in period, CTR hadto meet eligibility for randomization, in-cluding tolerance of at least 500 mgmetformin daily, performance of SMBGat least three times daily, maintenanceof HbA1c levels between 6.7 and 8.0%(50 and 64mmol/mol), 7–10-day use ofHolter devices, and successful 7–10-dayuse of CGM defined by a minimum1,836 glucose values (85% of a goal5-min CGM measurement for 7 days[2,160 total glucose values]). If CTR didnot reach the goal 2,160 CGM measure-ments, they continued CGM to achievethe goal or were vetted for extenuat-ing circumstances as long as theyachieved 85% of goal CGM readings be-fore randomization.

Randomized Treatment PeriodParticipants were randomized usingelectronic block randomization fromthe Data Coordinating Center at visit 3.Randomized participants continuedwith carefully supervised basal andmealtime therapy but were assigned ei-ther to continue BBI as in the run-in pe-riod or to replace the mealtime RAAwith mealtime dosing of the GLP-1RAexenatide (Byetta) using a pen injectorwhile continuing basal insulin glargine(GLIPULIN). Injections of either RAA orexenatide were generally taken just be-fore all significant meals, but the timingcould be adjusted according to partici-pant preferences and glycemic patterns.A total daily dose of up to 20 mg exena-tide was allowed, with distribution of 5-or 10-mg doses according to participant

needs. In general, twice-daily or threetimes daily dosing was used with bothregimens, depending on meal patterns.The investigators made titration deci-sions for basal and RAA insulin and forexenatide at all clinical visits as well asneeded with supplementary telephonecontact or electronic communications.The goal of titration was HbA1c withina range of 6.7–7.3% (50–56 mmol/mol).Point-of-care HbA1c was measured ateach clinic visit for clinical glycemicmanagement.

Clinical MeasurementsPostrandomization visits occurred atbaseline; 10 days; and 4, 12, 13, 19, 25,and 26 weeks (Supplementary Table 2).Clinical outcomes (blood pressure, heartrate, body weight, BMI, hip-waist circum-ference, adherence, and adverse events)were ascertained at baseline and 13 and26 weeks. Baseline and 13- and 26-weekvisits included downloads of SMBG, CGM,and Holter data as well as Core Laboratorymeasurement of HbA1c. Additional sam-ples for Core Laboratory measurementof albumin/creatinine ratio, creatinine, al-anine aminotransferase, inflammatorymarkers (levels of serum amyloid A[SAA], C-reactive protein [CRP], IL-6, andurinary 8-iso-prostaglandin F2a [8-iso-PGF2a] proteins) were also drawn atbaseline and 13 and 26 weeks. Samplesfor assessments of adiponectin, 1,5-anhydroglucitol (1,5-AG), HDL proteo-mics, and serum metabolomics wereperformed on a 60-participant subset(30 per group) of the cohort.

ECG Monitoring MeasurementsDuring each CGM monitoring period, acontinuous Holter ECGmonitor was alsoplaced on each participant for 7–10days. Each participant changed electro-des and batteries as needed and thenreturned the recorder for downloadingat the research site after completion. AllHolter datawere uploaded and processedbyMedicomp, Inc. Holter monitoring wasperformed together with CGM monitor-ing at prerandomization, 11–13 weeks,and 24–26 weeks. Other than the cor-rected QT interval (QTc), which was mea-sured manually, all arrhythmias werequantified by Medicomp algorithms andverified by the Holter ECG Reading andAnalysis Center at theUniversity ofWash-ington.

Processed Holter data were used toidentify “panic” and “alert” rhythms. Panic

rhythms were defined a priori as 1) com-plete heart block, 2) asystole (sinuspauses) of$5 s, 3) sustained ventriculartachycardia or wide complex tachycardiawith rate .150 beats/min (bpm) andduration $30 beats, or 4) ventricular fi-brillation. With any panic rhythm, thesite was notified, and further evaluationor treatment was at the discretion of theprimary investigator or the participant’sphysicians.

Alert rhythms were defined a priori as1) sustained supraventricular tachycar-dia (SVT) with rate$150 bpm and dura-tion$30 s; 2) sustained atrial fibrillationwith rate$150 bpm and duration$30 s;3) sustained sinus tachycardia with rate$200 bpm and duration $5 min; 4)sustained, marked bradycardia (sinus,idioventricular, or atrial fibrillation)with rate#35 bpm and duration$5 min;5) second-degree atrioventricular (AV)block types 1 and 2 or high-grade AVblock, 6) pauses with duration $3 and,5 s; and 7) nonsustained ventriculartachycardia (NSVT) or wide complextachycardia with rate $150 bpm andduration $5 beats and ,30 beats.

Prespecified Holter-assessed endpoints were 1) premature ventricularcomplexes (expressed as percent totalbeats per day); 2) premature supraven-tricular complexes (expressed as percenttotal beats per day); 3) QTc measuredonce daily using calipers when the heartrate was 60–100 bpm and the recordingfree of artifacts, with QTc calculatedbased on Bazett’s formula using the aver-ageheart rateover the preceding 10 s andexpressed as mean for all days of Holterrecording performed at the prerandom-ization, 11–13-week, and 24–26-week vis-its; 4) NSVT (expressed as number ofbeats per day and number of runs perday); 5) SVT (expressed as number ofbeats per day and number of runs perday); 6) sinus pauses .2.5 s (expressedas number of episodes per day); 7) atrialfibrillation burden (expressed as per-cent of beats per day); and 8) SVT bur-den (expressed as percent of beats perday).

Safety MeasurementsAt each visit and phone contact (Supple-mentary Table 2), CTR and randomizedparticipants were questioned to elicitreportable adverse events. Hypoglyce-mic events were identified by symptomsreported and review of diaries and

1560 Prandial Insulin Versus GLP-1 Agonist in T2D Diabetes Care Volume 38, August 2015

meter downloads. To further ensure safety,CGM data from the prerandomization,11–13-week, and final 24–26-week col-lections were analyzed. Predeterminedhypoglycemia detected by CGM was re-ported by the Data Coordinating Centerto the study site for the following: 1) aserious event, defined as $4% of read-ings ,40 mg/dL (2.2 mmol/L) with atleast 2,160 readings over a 7–10-day pe-riod, or 2) a significant event, defined as2 to ,4% of readings ,40 mg/dL (2.2mmol/L) or .5% of readings ,50 mg/dL(2.8mmol/L) with at least 2,160 readingsover a 7–10-day period. The investigatorcould reevaluate the participant statusthen make appropriate changes asneeded. Holter monitor recordingswere screened and sites notified ifpanic rhythms were identified. Seriousadverse events as well as unanticipatedadverse device or medication adverseevents were reported by the site inves-tigators, reviewed by the study safetyofficer, coded using the Medical Dic-tionary for Regulatory Activities, andreported to the pharmaceutical compa-nies for U.S. Food and Drug Administra-tion–mandated surveillance process.Safety data also were summarized andreported to the Data and Safety Moni-toring Board.

Statistical AnalysisA complete description of all the out-comes in FLAT-SUGAR are listed in hier-archical order in Table 1.

Primary Outcome

Change in the coefficient of variation(CV) of glucose values by CGM was theprimary outcome for this trial. The CV ofthe CGM is the SD of the glucose valuesdivided by themean. To estimate a sam-ple size for FLAT-SUGAR, unpublisheddata were taken from a previous rando-mized clinical trial of CGM (JDRF Contin-uous Glucose Sensor Trial) in participantswith type 1 diabetes (R. Beck, personalcommunication). The control group forthe JDRF trial (n = 58 adults aged .25years) had a baseline HbA1c of 6.7–8.0%(50–64 mmol/mol) and wore a maskedCGM at baseline and 6 months. At base-line, the group mean CV CGM value, cal-culated as the mean of individual meanCV CGM values, was 38, with an SD of 8and similar values at baseline. Values fol-lowed an approximately normal distribu-tion. Assuming a two-sample two-tailed ttest with a type I error of 0.05, a samplesize of 110 participants (55 per group)would give 90% power to detect a differ-ence of a mean change of 5 CV (SD 8)units between the control and treatmentgroups. In addition to the conventionaltwo-sample t test, an ANCOVA model

adjusting for the baseline value and clinicalsite was performed. If the residual val-ues from the ANCOVA indicatednonnormality, a Wilcoxon rank sumtest was used. To account for potentialdropouts and noncompliance with CGMuse, this value was increased by ;20to a sample size of 130 (65 per group).This sample size covers a range of pos-sible estimates in differences of CV of 3–6 and SD of CV of 4–8. With a power of85%, the sample size is 92 (108 allowingfor 15% dropout), and with a power of80%, the sample size is 82 (96 allowingfor 15% dropout). Analyses followed theintention-to-treat principle.

Secondary Outcomes

Because there is no gold standard to as-sess glycemic variability, especially in in-dividuals with type 2 diabetes, we alsoexamined other reported indices, in-cluding 1) relative improvement frombaseline of at least 20% of CV CGM; 2)percent of CGM readings 70–180 mg/dL(3.9–10.0 mmol/L), ,70 mg/dL (#3.9mmol/L) (hypoglycemia), and .180mg/dL (10.0 mmol/L) (hyperglycemia);3) SD; 4) mean amplitude of glycemicexcursions (MAGE); 5) continuous over-all net glycemic action (CONGA); 6)mean of daily differences (MODD); 7)interquartile range; and 8) mean glu-cose values at 48 h.

Table 1—Primary and secondary outcomes

Outcome hierarchy Measurement class Measurement type (source) Outcome measure

Primary outcome Glycemic control Glycemic variability (CGM) c CV CGM

Secondary outcomes Glycemic variability Glycemic variability (CGM) c SD in glucose levelsc MAGEc CONGAc MODD

Other glycemic indices c % glucose within 71–180 mg/dL by CGMc Hypoglycemia (clinical)

Biomarkers Systemic inflammation (serum) c IL-6c CRPc SAA

Oxidative stress (urine) c 8-iso-PGF2aDiabetic kidney disease (urine) c Albumin/creatinine ratio

Weight control Body weight (vital signs) c Body weightc BMI

Rhythm disturbance Arrhythmias (Holter monitor) c QTcBy randomized group

c ArrhythmiasBy randomized group:Supraventricular arrhythmiasVentricular arrhythmias

Tertiary outcomes Glycemia Hyperglycemia (serum) c 1,5-AGAdipocyte biology Adipokine (serum) c AdiponectinPlasma lipoprotein HDL proteins (plasma) c HDL proteomicsSystemic metabolism Metabolites (serum) c Metabolomics

care.diabetesjournals.org The FLAT-SUGAR Trial Investigators 1561

RESULTS

Between August 2012 and January 2014,the 12 clinical sites screened 255 indi-viduals, and 131 eligible candidates en-tered the run-in period (SupplementaryTable 3). Of these, 102 were eligibleafter the run-in. The leading causes ofineligibility were HbA1c out of range(n = 101), dropout (n = 12), and C-peptidelevel out of range (n = 17). All 102 par-ticipants agreed to be randomized toeither BBI or GLIPULIN. Baseline char-acteristics of randomized participantswere balanced between treatmentgroups (Table 2) and closely resembledthose of the ACCORD cohort. Sixty-three percent were male, 81% were Cau-casian, mean age was 62 years, and 33%had a prior CVD event. At randomization,mean BMI was 34 kg/m2, blood pressure130/73mmHg, HbA1c 7.9% (63mmol/mol),and creatinine 79.56 mmol/L. Micro-albuminuria was present in 18% andmacroalbuminuria in 3% of participantsat randomization.

CGM Data at BaselineA graphic representation of the 24-hmean of the means for all randomizedparticipants is shown in Fig. 1. The low-est mean value occurred between 5:00and 10:00 A.M., with a prominent incre-ment after the first meal and smallerincrements later in the day. All partic-ipants completed the minimum num-ber of CGM measurements, although10 did not achieve the goal CGM of2,160 readings. The mean value forCV CGM at baseline in the BBI group(n = 49 completed) was 30.1 6 6.0and in the GLIPULIN group (n = 52),31.9 6 6.1 (P not significant for thedifference). Only 2% of all CGM read-ings were ,70 mg/dL (3.9 mmol/L);;23% were .180 mg/dL (10 mmol/L).Mean values of MAGE, CONGA, andMODD are shown in Table 3. These sec-ondary glycemic outcomes were com-pared with the final levels at the endof the trial.

ECG and Holter Monitoring Values atBaselineA total of 102 participants had a baselinecontinuous Holter monitor placed beforerandomization for a meanmonitoring du-ration of 8 days per participant. The totalnumber of rhythm disturbances was rel-atively few during the monitoring period.

No panic rhythmswere detected. Six alertrhythms were identified during the pre-randomization Holter recordings usinga priori definitions. For participantsassigned to the BBI group, detectedrhythms were sustained atrial fibrillation(one occurrence in one participant) and

NSVT (one occurrence in three partici-pants). For participants assigned to theGLIPULIN group, detected rhythms weresecond-degree AV block (one occurrencein one participant) and a sinus pause be-tween 3 and 5 s (one occurrence in oneparticipant).

Table 2—Characteristics of participants at randomization

VariableGLIPULIN(n = 52)

BBI(n = 50)

Total(n = 102)

ACCORDglycemia trial(n = 10,251)

Age (years) 62 (8) 63 (7) 62 (8) 62 (7)

Female sex 31 44 37 39

Median duration ofdiabetes (years) 15 (9) 16 (7) 15 (8) 10

Previous cardiovascularevent 39 28 33 35

Previous congestiveheart failure 8 0 4 5

Race or ethnic groupWhite non-Hispanic 83 80 81 63Black non-Hispanic 12 14 13 19Hispanic 2 6 4 7Other 4 0 2 11

Weight (kg) 101.3 (14.2) 99.7 (17.8) 100.5 (16.0) 93.4

BMI (kg/m2) 34.2 (5.0) 33.7 (5.1) 33.9 (5.1) 32.2 (5.5)

Waist circumference (cm) 113.9 (16.6) 112.8 (14.1) 113.4 (15.4) 106.8

Blood pressure (mmHg)Systolic 131.7 (13.6) 129.1 (15.2) 130.4 (14.4) 136.4Diastolic 73.5 (11.3) 73.1 (9.8) 73.3 (10.5) 74.9

Heart rate (bpm) 70.5 (10.5) 72.4 (10.4) 71.4 (10.4) 72.7

BloodHbA1c (mmol/mol) 64 (4) 63 (3) 63 (4) 67 (12)HbA1c (%) 8.0 (0.4) 7.9 (0.3) 7.9 (0.4) 8.3 (1.1)C-peptide (nmol/L) 0.80 (0.07) 0.73 (0.10) 0.77 (0.07) NACreatinine (mmol/L) 79.56 (17.68) 79.56 (17.68) 79.56 (17.68) 79.56Alanine aminotransferase

(mkat/L) 37.6 (22.8) 32.4 (14.8) 35.0 (19.3) NA

UrineAlbumin/creatinine ratio

(mg/g) 3.95 (7.83) 4.52 (16.94) 4.23 (13.05) NA,3.4 75 84 80 693.4–34 21 14 18 25.34 4 2 3 7

Glucose-lowering medications(prescreening)

Insulin 100 100 100 35Basal insulin only 27 24 26 NAPrandial insulin only 2 6 4 NABasal + one or more

prandial injections 62 58 60 NAPremixed insulin one or

two injections 4 6 5 NAMetformin 89 90 89 60Secretagogue 21 26 24 51Thiazolidinedione 6 2 4 19GLP-1RA 0 8 4 0DPP-4 inhibitor 2 4 3 0

CV CGM 31.9 (6.1) 30.0 (6.1) 31.0 (6.1) NA

Data are mean (SD) or %. DPP-4, dipeptidyl peptidase-4; NA, not available.

1562 Prandial Insulin Versus GLP-1 Agonist in T2D Diabetes Care Volume 38, August 2015

CONCLUSIONS

The FLAT-SUGAR study differs from pre-vious studies of glycemic variability inseveral important ways, including its op-erational structure, the therapeuticmethods used, and the population stud-ied. Although an investigator-initiatedstudy, FLAT-SUGAR has a complex de-sign, involves geographically dispersedclinical centers, and uses centralized fa-cilities for operational management,data handling, laboratory measure-ments, and drug and device supply andsupport (Supplementary Fig. 1). Finan-cial support was provided primarilyby a single commercial entity, with sig-nificant additional amounts provided

from two other companies. Severalother companies provided support inthe form of in-kind donations of studydrugs and equipment (details providedin DUALITY OF INTEREST).

It became clear in the 1990s that gly-cemic control as assessed by HbA1c

was amajor risk factor for microvascularand neuropathic complications (1,2).However, over the next two decades,there has been support for the conceptthat these results arise from factorsother than the mean glucose alone,which is captured by HbA1c (12,17).Given the evidence that both hypergly-cemia and hypoglycemia activate in-flammation and oxidative stress, we

hypothesized that the variability ofglucose could be another risk factorfor diabetes-related complications. Thetriumvirate of glucose variability, hyper-glycemia, and hypoglycemia, with inter-actions among these components, hasbeen identified and reviewed byMonnieret al. (34). In Fig. 2, we summarize themechanisms potentially linking these as-pects of glycemic control to oxidativestress and inflammatory mechanisms.

Although it is agreed that the CV maynot completely capture glucose varia-bility, this variable was identified asthemost practical outcome for monitor-ing partly because we had previous datathat allowed for an estimate of samplesize for this study. Other measures ofglycemic variability were also includedin the analytic plan (Table 1). The pri-mary purpose for FLAT-SUGAR was todetermine whether separating glycemicvariability while maintaining similarHbA1c levels in a population of obesepatients with type 2 diabetes and athigh risk for or with known CVD is pos-sible. A positive study would lead us toperform a definitive study with a rele-vant clinical outcome. Additional pre-specified measures of outcome wereECG measurements collected by Holtermonitoring, albuminuria, and variousmeasures of inflammation and oxidativestress that may support a link betweenglycemic variability and tissue injury.

The study questions posed about thepotential to reduce glycemic variabilityand the effect of this effort on physio-logic and medical outcomes could nothave been tested without availabilityof the shorter-acting GLP-1RAs that mayblunt postprandial glycemic incrementsin type 2 diabetes more effectively thanprandial insulin or long-acting GLP-1s. Al-though important reductions of glycemicvariability have been shown with use oftwice-daily injections of exenatide (35),FLAT-SUGAR tested the potential for aneven-greater reduction of variability withinjections of exenatide at low doses up tothree times daily with meals. Similarly,use of currently available devices forCGM allows for collection of completedaily profiles to document the changesof glycemic profiles obtained by the twotreatment methods. Although most ofthe evidence supporting a role for glyce-mic variability in the pathogenesis of dia-betes complications has come, up to now,from epidemiologic studies or short-term

Figure 1—Graph of mean of means of 24-h CGM readings (for all 102 participants) beforerandomization in FLAT-SUGAR. Error curves represent the SD.

Table 3—Metrics of variability from CGM

Baseline (n = 102) mg/dL mmol/L

CV CGM 31.03 (6.07) d d

% CGM readings within 70–180 mg/dL(3.9–10 mmol/L) 74.32 (11.84) d d

Hypoglycemia (% readings ,70 mg/dL[3.9 mmol/L]) 2.49 (3.01) d d

Hyperglycemia (% readings .180 mg/dL[10 mmol/L]) 23.19 (12.13) d d

Mean glucose d 149.27 (17.25) 8.28 (0.96)

SD d 46.37 (10.65) 2.57 (0.59)

MAGE d 85.93 (19.05) 4.77 (1.06)

CONGA d 59.12 (15.89) 3.28 (0.88)

MODD d 45.89 (12.52) 2.55 (0.69)

Interquartile range d 59.97 (14.48) 3.33 (0.80)

Data are mean (SD). Mean glucose is the average of 48-h CGM values for each patient.

care.diabetesjournals.org The FLAT-SUGAR Trial Investigators 1563

physiologic studies of surrogate measures,FLAT-SUGARmayprovidemoredirect inter-ventional evidence. If, as intended, equiva-lent HbA1c levels are attained with BBI andGLIPULIN treatment, the effects of reducedglycemic variability may be assessed, al-though direct effects of the exenatide orprandial insulin treatments cannot be ex-cluded. The array of biomarkers includedin the study will provide insights into a va-riety of postulated mechanisms.The population selected for FLAT-

SUGAR is a high-risk subgroup of peoplewith type 2 diabetes. That the partici-pants needed both basal insulin and ad-ditional prandial treatment ensures arelatively long duration of diabetes,and clinical markers of CVD risk are ap-parent at randomization. Comparison ofbaseline characteristics of the FLAT-SUGAR and ACCORD populations atbaseline confirms considerable similar-ity between the two. Because microal-buminuria was present in 18% and aprior CVD event had occurred in 33%of those randomized in FLAT-SUGAR,

assessment of changes in albuminuriaand occurrence of ECG abnormalitieswill be of great interest. Demonstrationof differences in these measures be-tween the groups assigned to BBI andGLIPULIN could provide justificationfor a larger and longer study includingthese and other medical end points.

Despite these novel features, FLAT-SUGAR has important limitations. A truecontrol group is not feasible. This limita-tion introduces the difficulty of distin-guishing between the effects of reducedglycemic variability and nonglycemic ef-fects of exenatide or prandial insulin(36). There could be direct tissue effectsor ones resulting from nonglycemic met-abolic effects or from weight changes. Fi-nally, the study is, by necessity, unblinded.

Because the study requires an inten-sive treatment schedule with frequentvisits, multiple interactions with highlyspecialized staff, and frequent measuresof outcomes including both CGM andHolter monitoring, conclusions from theresults cannot be reliably generalized to a

broader population of less motivated orless carefully monitored people. Addition-ally, the protocol is not consistent withusual clinical practice in adding a GLP-1RA after BBI therapy, which has proveninsufficiently effective. However, the goalis not to replicate practice but to test thehypothesis that glucose variability can bedifferentiated in this population.

In summary, FLAT-SUGAR is a proof-of-concept study aiming to verify theability of a novel combination therapyregimen of basal insulin with a short-acting GLP-1RA to decrease glycemicvariability and to then investigatewhether this decrease in variability cor-relates with biochemical and clinicalpredictors of adverse medical outcomes.We successfully recruited, as intended, atrial population that resembles that of theACCORD study population in which inten-sive glycemic therapy proved tobepoten-tially hazardous. The randomized groupsin FLAT-SUGAR were evenly matched.The number of participants recruited pro-vides sufficient power to answer the pri-mary study question and to address anumber of secondary questions of inter-est. Validation of the treatment approachtested in FLAT-SUGAR and identificationof the best clinical variables for furtherinvestigation could justify and assist inplanning a larger medical outcomes trial.

Appendix

Writing Committee. Jeffrey L. Probstfield, Di-vision of Cardiology, University of Washington(co-chair, co-principal investigator); Irl Hirsch,Diabetes Care Center, University of Washington(co-principal investigator); Kevin O’Brien, Divisionof Cardiology, University of Washington; BarryDavis, School of Public Health, University of TexasHealth Science Center at Houston; Richard Ber-genstal, International Diabetes Center, Park Nic-ollet; Connie Kingry, Division of Cardiology,University of Washington; Dori Khakpour, Diabe-tes Care Center, University of Washington; SarahPressel, School of Public Health, University ofTexas Health Science Center at Houston; KelleyR. Branch, Division of Cardiology, University ofWashington; and Matthew Riddle, Section of Di-abetes, Oregon Health & Science University(co-chair).Funding. B.D. received honoraria from theHelmsley Charitable Trust.Duality of Interest. This study is an investiga-tor-initiated clinical trial primarily funded by agrant from Sanofi. A contract was developedbetween Sanofi and the Clinical Trials ServiceUnit of the University of Washington, whichbecame the academic sponsor of the study.Supplemental funding was provided later in the

Figure 2—Selected, potential biological consequences of hyper- and hypoglycemia assessed inFLAT-SUGAR. Hyperglycemia (and/or fatty acids) may uncouple mitochondrial respiration, lead-ing to 1) increased oxidative stress, as assessed by measuring urinary 8-iso-PGF2a, a product ofarachidonic acid oxidation; 2) increased inflammatory activation, as assessed by measuringplasma levels of IL-6, CRP, and SAA resulting from translocation to the nucleus of the transcrip-tion factor nuclear factor-kB; and 3) secondary activation of additional CRP and SAA expressionby circulating IL-6. Hypoglycemia may also stimulate inflammatory activation secondarily, lead-ing to an adrenergic counterregulatory response resulting in epinephrine- and/or norepineph-rine-mediated stimulation of b-receptor–mediated 4) lipolysis, with increased intracellular fattyacids and inflammation also assessed through plasma IL-6, CRP, and SAA levels; 5) direct stim-ulation of arrhythmias, as assessed by Holter monitoring; and 6) effects on multiple (e.g.,potassium [IKr], chloride [Cl2], calcium [Ca2+]) ion channels, resulting in an increase in theQTc, which predisposes to arrhythmia and is assessed by Holter monitoring. IkB, inhibitor of kB.

1564 Prandial Insulin Versus GLP-1 Agonist in T2D Diabetes Care Volume 38, August 2015

study by grants from Bristol-Myers Squibb-AstraZeneca and Sanofi. In-kind donationswere provided for the study by several organ-izations, including Sanofi (donated insulin glargine[Lantus] and insulin glulisine [Apidra] and pur-chased insulin aspart [NovoLog]), Eli Lilly (donatedinsulin lispro [Humalog]), Amylin/Bristol-MyersSquibb-AstraZeneca (donated exenatide [Byetta]),Bayer (donatedContour glucosemeters, strips, andsupplies), BectonDickinson (donated pen needles),Dexcom (donated the various components of theCGM devices [SEVEN PLUS and G4]), and Medi-comp (donated Holter monitoring equipment andsupported reading of the recordings).

J.L.P. reports honoraria for consulting fromSanofi and research grant support from Sanofi,Bristol-Myers Squibb-AstraZeneca, Eli Lilly, andBayer; these dualities of interest have beenreviewed by the University of Washington. I.H.reports receiving honoraria for consulting fromRoche, Abbott Diabetes Care, and Becton Dick-inson and receiving grants from Sanofi andHalozyme. K.O. reports having received grantsupport from Sanofi. B.D. reports having re-ceived grant support from Sanofi. R.B. reportsreceiving honoraria for consulting and lecturesfrom Abbott Diabetes Care, Bayer, BoehringerIngelheim, Bristol-Myers Squibb-AstraZeneca,Calibra, Eli Lilly, Halozyme, Hygieia, Johnson &Johnson, Medtronic, Novo Nordisk, Roche,Sanofi, Takeda, and Valeritas. C.K. reports hav-ing received grant support from Sanofi andBristol-Myers Squibb-AstraZeneca. D.K., S.P., andK.R.B. report having received grant support fromSanofi. M.R. reports having received grant sup-port from Sanofi, AstraZeneca, Eli Lilly, andNovo Nordisk; honoraria for speaking fromSanofi; and honoraria for consulting from Sanofi,AstraZeneca, Eli Lilly, Elcelyx, Valeritas, and Bio-del; these dualities of interest have been re-viewed and managed by Oregon Health &Science University. No other potential conflictsof interest relevant to this article were reported.Author Contributions. J.L.P., I.H., and C.K.contributed to the study concept and design;obtaining of funding; provision of administra-tive, technical, or logistical support; provision ofmaterials, patients, or resources; data collec-tion; data analysis and interpretation; first draftof the manuscript; and revision and finalapproval of the manuscript. K.O. contributedto the study concept and design; obtaining offunding; provision of administrative, technical,or logistical support; data collection; dataanalysis and interpretation; and revision andfinal approval of the manuscript. B.D. contrib-uted to the study concept and design; obtainingof funding; provision of administrative, techni-cal, or logistical support; provision of materials,patients, or resources; data analysis and in-terpretation; statistical analysis; first draft ofthe manuscript; and revision and final approvalof the manuscript. R.B. contributed to theprovision of administrative, technical, or logis-tical support; provision of materials, patients, orresources; and revision and final approval of themanuscript. D.K. contributed to the studyconcept and design; provision of administrative,technical, or logistical support; provision ofmaterials, patients, or resources; data collec-tion; data analysis and interpretation; first draftof the manuscript; and revision and final

approval of the manuscript; S.P. contributedto the provision of administrative, technical, orlogistical support; provision of materials, pa-tients, and resources; data collection; dataanalysis and interpretation; statistical analysis;first draft of the manuscript; and revision andfinal approval of the manuscript. K.R.B. contrib-uted to the data analysis and interpretation andrevision and final approval of the manuscript.M.R. contributed to the study concept anddesign, first draft of themanuscript, and revisionand final approval of the manuscript. J.L.P. is theguarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrity of the data and theaccuracy of the data analysis.

References1. The Diabetes Control and Complications TrialResearch Group. The effect of intensive treat-ment of diabetes on the development andprogression of long-term complications in insu-lin-dependent diabetes mellitus. N Engl J Med1993;329:977–9862. UK Prospective Diabetes Study (UKPDS)Group. Intensive blood-glucose control with sul-phonylureas or insulin compared with conven-tional treatment and risk of complications inpatients with type 2 diabetes (UKPDS 33). Lan-cet 1998;352:837–8533. American Diabetes Association. Standards ofmedical care in diabetesd2014. Diabetes Care2014;37(Suppl. 1):S14–S804. Rodbard HW, Blonde L, Braithwaite SS, et al.;AACE Diabetes Mellitus Clinical Practice Guide-lines Task Force. American Association of Clini-cal Endocrinologists medical guidelines forclinical practice for themanagement of diabetesmellitus [published correction appears in En-docr Pract 2008;14:802–803]. Endocr Pract2007;13(Suppl. 1):1–685. Duckworth W, Abraira C, Moritz T, et al.;VADT Investigators. Glucose control and vascu-lar complications in veterans with type 2 diabe-tes. N Engl J Med 2009;360:129–1396. Gerstein HC, Miller ME, Byington RP, et al.;Action to Control Cardiovascular Risk in Diabe-tes Study Group. Effects of intensive glucoselowering in type 2 diabetes. N Engl J Med2008;358:2545–25597. Patel A, MacMahon S, Chalmers J, et al.;ADVANCE Collaborative Group. Intensive bloodglucose control and vascular outcomes in pa-tients with type 2 diabetes. N Engl J Med2008;358:2560–25728. Esposito K, Nappo F, Marfella R, et al. Inflam-matory cytokine concentrations are acutely in-creased by hyperglycemia in humans: role ofoxidative stress. Circulation 2002;106:2067–20729. Razavi Nematollahi L, Kitabchi AE, Stentz FB,et al. Proinflammatory cytokines in response toinsulin-induced hypoglycemic stress in healthysubjects [published correction appears in Me-tabolism 2009;58:1046]. Metabolism 2009;58:443–44810. Brownlee M. The pathobiology of diabeticcomplications: a unifying mechanism. Diabetes2005;54:1615–162511. Ceriello A, Esposito K, Piconi L, et al. Oscil-lating glucose is more deleterious to endothelial

function and oxidative stress than mean glucosein normal and type 2 diabetic patients. Diabetes2008;57:1349–135412. El-Osta A, Brasacchio D, Yao D, et al. Tran-sient high glucose causes persistent epigeneticchanges and altered gene expression duringsubsequent normoglycemia. J Exp Med 2008;205:2409–241713. Esposito K, Ciotola M, Carleo D, et al. Post-meal glucose peaks at home associate with ca-rotid intima-media thickness in type 2 diabetes.J Clin Endocrinol Metab 2008;93:1345–135014. Chow E, Bernjak A, Williams S, et al. Risk ofcardiac arrhythmias during hypoglycemia in pa-tients with type 2 diabetes and cardiovascularrisk. Diabetes 2014;63:1738–174715. Desouza C, Salazar H, Cheong B, Murgo J,Fonseca V. Association of hypoglycemia and car-diac ischemia: a study based on continuousmonitoring. Diabetes Care 2003;26:1485–148916. Nordin C. The proarrhythmic effect of hy-poglycemia: evidence for increased risk fromischemia and bradycardia. Acta Diabetol 2014;51:5–1417. BrownleeM, Hirsch IB. Glycemic variability:a hemoglobin A1c-independent risk factor fordiabetic complications. JAMA 2006;295:1707–170818. Standl E, Schnell O, Ceriello A. Postprandialhyperglycemia and glycemic variability: shouldwe care? Diabetes Care 2011;34(Suppl. 2):S120–S12719. Holman RR, Farmer AJ, Davies MJ, et al.; 4-TStudy Group. Three-year efficacy of complex in-sulin regimens in type 2 diabetes. N Engl J Med2009;361:1736–174720. Riddle MC, Rosenstock J, Vlajnic A, Gao L.Randomized, 1-year comparison of three waysto initiate and advance insulin for type 2 diabe-tes: twice-daily premixed insulin versus basalinsulin with either basal-plus one prandial insu-lin or basal-bolus up to three prandial injections.Diabetes Obes Metab 2014;16:396–40221. Kolterman OG, Kim DD, Shen L, et al. Phar-macokinetics, pharmacodynamics, and safety ofexenatide in patients with type 2 diabetes mel-litus. Am J Health Syst Pharm 2005;62:173–18122. Cervera A, Wajcberg E, Sriwijitkamol A,et al. Mechanism of action of exenatide to re-duce postprandial hyperglycemia in type 2 di-abetes. Am J Physiol Endocrinol Metab 2008;294:E846–E85223. Kapitza C, Forst T, Coester H-V, Poitiers F,Ruus P, Hincelin-Mery A. Pharmacodynamiccharacteristics of lixisenatide once daily versusliraglutide once daily in patients with type 2 di-abetes insufficiently controlled on metformin.Diabetes Obes Metab 2013;15:642–64924. Meier JJ. GLP-1 receptor agonists for indi-vidualized treatment of type 2 diabetes melli-tus. Nat Rev Endocrinol 2012;8:728–74225. Riddle MC, Drucker DJ. Emerging therapiesmimicking the effects of amylin and glucagon-like peptide 1. Diabetes Care 2006;29:435–44926. Buse J, Bergenstal R, Glass LC, et al. Use oftwice-daily exenatide in basal insulin-treatedpatients with type 2 diabetes: a randomized,controlled trial. Ann Intern Med 2011;154:103–11227. Diamant M, Nauck MA, Shaginian R, et al.;4B Study Group. Glucagon-like peptide-1 recep-tor agonist or bolus insulin with optimized basal

care.diabetesjournals.org The FLAT-SUGAR Trial Investigators 1565

insulin in diabetes. Diabetes Care 2014;37:2763–277328. Riddle MC, Aronson R, Home P, et al. Add-ing once-daily lixisenatide for type 2 diabetesinadequately controlled by established basal in-sulin: a 24-week, randomized, placebo-controlledcomparison (GetGoal-L). Diabetes Care 2013;36:2489–249629. Riddle MC, Forst T, Aronson R, et al. Addingonce-daily lixisenatide for type 2 diabetes inad-equately controlled with newly initiated andcontinuously titrated basal insulin glargine:a 24-week, randomized, placebo-controlledstudy (GetGoal-Duo 1). Diabetes Care 2013;36:2497–250330. Hirsch IB, Armstrong D, Bergenstal RM,et al. Clinical application of emerging sensor

technologies in diabetes management: con-sensus guidelines for continuous glucose mon-itoring (CGM). Diabetes Technol Ther 2008;10:232–24431. Rodbard D. Interpretation of continuousglucose monitoring data: glycemic variabilityand quality of glycemic control. Diabetes Tech-nol Ther 2009;11(Suppl. 1):S55–S6732. Mazze RS, Strock E, Borgman S, Wesley D,Stout P, Racchini J. Evaluating the accuracy,reliability, and clinical applicability of continu-ous glucose monitoring (CGM): is CGM readyfor real time? Diabetes Technol Ther 2009;11:11–1833. Bergenstal RM, Ahmann AJ, Bailey T, et al.Recommendations for standardizing glucose re-porting and analysis to optimize clinical decision

making in diabetes: the Ambulatory GlucoseProfile (AGP). Diabetes Technol Ther 2013;15:198–21134. Monnier L, Colette C, Owens D. The glyce-mic triumvirate and diabetic complications: isthe whole greater than the sum of its compo-nent parts? Diabetes Res Clin Pract 2012;95:303–31135. McCall AL, Cox DJ, Brodows R, Crean J,Johns D, Kovatchev B. Reduced daily risk of gly-cemic variability: comparison of exenatide withinsulin glargine. Diabetes Technol Ther 2009;11:339–34436. Monnier L, Colette C,Mas E, et al. Regulationof oxidative stress by glycaemic control: evidencefor an independent inhibitory effect of insulintherapy. Diabetologia 2010;53:562–571

1566 Prandial Insulin Versus GLP-1 Agonist in T2D Diabetes Care Volume 38, August 2015